Advanced Natural Language Processing and Deep Learning
In this course students will learn to apply modern state-of-the-art solutions for natural language processing problems. We go beyond simple classification tasks, and tackle more advanced types of tasks, like generation and structured prediction.
This course covers advanced natural language processing tasks, models, and setups. The course builds on the Introduction to Natural Language Processing (NLP) and Deep Learning course (BSSEYEP1KU). We will focus on more advanced tasks, including structured prediction (such as finding relations between words), text generation and multi-task learning, all using modern state-of-the-art language models. Furthermore, we will address low resource scenarios, and the students will learn to build NLP models when no training data for the language or language type of interest is available.
The student should be able to implement algorithms in python. The student must have taken the Introduction to Natural Language Processing (NLP) and Deep Learning course (BSSEYEP1KU) offered in the BSc Data Science program, or an equivalent course covering at least classification and sequence labeling in NLP.
If the student did not attend an NLP course at ITU or elsewhere, they should request access to the LearnIT page of the Introduction to Natural Language Processing (NLP) and Deep Learning course (BSSEYEP1KU) course and study the materials. Specifically, they should focus on Chapters 4-8 of the textbook (https://web.stanford.edu/~jurafsky/slp3/old_dec21/indexdec21.html).
Intended learning outcomes
After the course, the student should be able to:
- Summarize recent research in NLP
- Present recent research in NLP
- Evaluate and compare a variety of NLP models
- Recommend accurate solutions for a wide range of NLP tasks
- Design and build state-of-the-art solutions for a wide range of NLP tasks and setups
- Formulate a relevant research question, embedded in current literature
- Report the outcomes of a research project, answering a research question
- A series of lectures with corresponding assignments
- Group presentation of a relevant research paper
- Final group (research) project with submitted report
Presentation of a research paper related to one of the covered topics.
The pedagogical purpose of the mandatory activities is to learn to interpret recent research papers and clearly communicate their content and takeaways.
The students will receive oral formative feedback after the presentation from a teacher.
If the students fail to hand in or get a “not approved” of the mandatory activity they will get a second attempt in a later lecture.
The student will receive the grade NA (not approved) at the ordinary exam, if the mandatory activities are not approved and the student will use an exam attempt.
The main textbook is the third edition of Speech and Language Processing (available for free at: https://web.stanford.edu/~jurafsky/slp3/old_dec21/indexdec21.html). In addition we will use research papers, which will be listed on the LearnIt page
Student Activity BudgetEstimated distribution of learning activities for the typical student
- Preparation for lectures and exercises: 16%
- Lectures: 8%
- Exercises: 20%
- Project work, supervision included: 40%
- Exam with preparation: 16%
Ordinary examExam type:
D: Submission of written work with following oral, Internal (7-point scale)
D2G: Submission for groups with following oral exam supplemented by the submission. Shared responsibility for the report.
Submissions will be based on a group (3-4 students) research project. You have to upload a research paper describing the outcomes of your research project (pdf), with corresponding code in a git repository. More detailed information is available on LearnIt.
Mixed exam 1 : Individual and joint student presentation followed by an individual and a group dialogue. The students make a joint presentation followed by a group dialogue. Subsequently the students are having individual examination with presentation and / or dialogue with the supervisor and external examiner while the rest of the group is outside the room.
Time and dateOrdinary Exam - submission Thu, 21 Dec 2023, 08:00 - 14:00
Ordinary Exam Tue, 9 Jan 2024, 09:00 - 21:00
Ordinary Exam Wed, 10 Jan 2024, 09:00 - 21:00
Ordinary Exam Thu, 11 Jan 2024, 09:00 - 21:00